Serologic Biomarkers of Progression Toward Diagnosis of Rheumatoid Arthritis in Active Component Military Personnel

Objective: To identify a panel of serum biomarkers that could specifically identify imminent cases of rheumatoid arthritis (RA) before diagnosis. Methods: Serum samples were collected at 4 time points from active component US military personnel, including 157 anti–citrullinated protein antibody (ACPA)–seropositive and 50 ACPA-seronegative RA subjects, 100 reactive arthritis (ReA) subjects, and 76 healthy controls. The cohorts were split into 2 phases, with samples tested on independent proteomic platforms for each phase. Classification models of RA diagnosis based on samples obtained within 6 months prior to diagnosis were developed both in univariate analyses and by multivariate random forest modeling of training sample sets and testing sample sets from each phase. Results: Increases in serum analytes, including C-reactive protein levels, serum amyloid A, and soluble programmed cell death 1 (PD-1), were observed in seropositive RA subjects at the time point closest to diagnosis, up to several years before diagnosis. Only a small fraction of RA subjects had levels above the 95th percentile of healthy control levels until the time period within 6 months of diagnosis. For classification of RA diagnosis using samples obtained within 6 months prior to diagnosis, soluble PD-1 provided superior specificity compared to ReA cases (>89%), with a sensitivity of 48% for RA classification. An 8-analyte model provided superior sensitivity (69%), with comparable specificity relative to ReA (>82%). Conclusion: Our findings demonstrate that imminent RA diagnosis could be classified with high specificity, relative to healthy controls and ReA cases, using a panel of cytokines measured in serum samples collected within 6 months before actual diagnosis.